# AI Apps vs Models

At Supervised, so far, you were able to build AI apps on some pre-defined models : 

<figure><img src="https://26047275-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fz7k4QvTbfR7aTAdgPmkt%2Fuploads%2FppsHFbd2BAPblry13In4%2FScreenshot%202023-06-11%20at%2012.47.55%20PM.png?alt=media&#x26;token=aae3b0d7-a3e1-4fb6-9cd7-7aa7ebd14a55" alt=""><figcaption></figcaption></figure>

You had the option to choose between GPT-4, GPT-3, Davinci 003 and Davinci 002. However, these are the foundational models of GPT that are very generalized in solving specific doubts.&#x20;

So, in the most layman sense, you can tell that a model is the foundation of any AI engine and the AI app is a tailored alteration of the model that works on a prompt defined by the makers.&#x20;

<figure><img src="https://26047275-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fz7k4QvTbfR7aTAdgPmkt%2Fuploads%2F2sKlZAGuHWecIfQ2Qfba%2FSttabot%20PP.png?alt=media&#x26;token=712e7357-3fde-4a79-bda0-c6447dd69f40" alt=""><figcaption></figcaption></figure>

Now, it might be clear to you that models (that are now available in Supervised 2.0), are basically a way to take more control your over AI apps, and tailor them around your own data set to make it unique and reliable.&#x20;